@inproceedings{song-etal-2025-instantly,
title = "Instantly Learning Preference Alignment via In-context {DPO}",
author = "Song, Feifan and
Fan, Yuxuan and
Zhang, Xin and
Wang, Peiyi and
Wang, Houfeng",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.8/",
doi = "10.18653/v1/2025.naacl-long.8",
pages = "161--178",
ISBN = "979-8-89176-189-6",
abstract = "Human Preference Alignment (HPA) can assist large language models (LLMs) to generate safe content. Due to the heavy cost of fine-tuning, tuning-free methods have emerged, typically modifying LLM decoding via post-processing. In this paper, we propose a novel and effective approach for HPA in a tuning-free way, named In-Context Direct Preference Optimization (ICDPO). We first rethink the derivation procedures of DPO, based on which we conversely build an instant scorer using the states of the LLM before and after ICL. It enables LLMs to both generate and select the well-aligned response, which is precisely estimated by the aforementioned instant scorer, thereby enhancing the final performance. ICDPO can be further enhanced with a two-stage retriever and an upgraded scorer. Extensive experiments show its effectiveness, particularly in outperforming multiple tuning-free baselines, even competitiveness with SFT and DPO. We also conduct detailed analyses to offer comprehensive insights into ICDPO."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="song-etal-2025-instantly">
<titleInfo>
<title>Instantly Learning Preference Alignment via In-context DPO</title>
</titleInfo>
<name type="personal">
<namePart type="given">Feifan</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peiyi</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Houfeng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>Human Preference Alignment (HPA) can assist large language models (LLMs) to generate safe content. Due to the heavy cost of fine-tuning, tuning-free methods have emerged, typically modifying LLM decoding via post-processing. In this paper, we propose a novel and effective approach for HPA in a tuning-free way, named In-Context Direct Preference Optimization (ICDPO). We first rethink the derivation procedures of DPO, based on which we conversely build an instant scorer using the states of the LLM before and after ICL. It enables LLMs to both generate and select the well-aligned response, which is precisely estimated by the aforementioned instant scorer, thereby enhancing the final performance. ICDPO can be further enhanced with a two-stage retriever and an upgraded scorer. Extensive experiments show its effectiveness, particularly in outperforming multiple tuning-free baselines, even competitiveness with SFT and DPO. We also conduct detailed analyses to offer comprehensive insights into ICDPO.</abstract>
<identifier type="citekey">song-etal-2025-instantly</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.8</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.8/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>161</start>
<end>178</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Instantly Learning Preference Alignment via In-context DPO
%A Song, Feifan
%A Fan, Yuxuan
%A Zhang, Xin
%A Wang, Peiyi
%A Wang, Houfeng
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F song-etal-2025-instantly
%X Human Preference Alignment (HPA) can assist large language models (LLMs) to generate safe content. Due to the heavy cost of fine-tuning, tuning-free methods have emerged, typically modifying LLM decoding via post-processing. In this paper, we propose a novel and effective approach for HPA in a tuning-free way, named In-Context Direct Preference Optimization (ICDPO). We first rethink the derivation procedures of DPO, based on which we conversely build an instant scorer using the states of the LLM before and after ICL. It enables LLMs to both generate and select the well-aligned response, which is precisely estimated by the aforementioned instant scorer, thereby enhancing the final performance. ICDPO can be further enhanced with a two-stage retriever and an upgraded scorer. Extensive experiments show its effectiveness, particularly in outperforming multiple tuning-free baselines, even competitiveness with SFT and DPO. We also conduct detailed analyses to offer comprehensive insights into ICDPO.
%R 10.18653/v1/2025.naacl-long.8
%U https://aclanthology.org/2025.naacl-long.8/
%U https://doi.org/10.18653/v1/2025.naacl-long.8
%P 161-178
Markdown (Informal)
[Instantly Learning Preference Alignment via In-context DPO](https://aclanthology.org/2025.naacl-long.8/) (Song et al., NAACL 2025)
ACL
- Feifan Song, Yuxuan Fan, Xin Zhang, Peiyi Wang, and Houfeng Wang. 2025. Instantly Learning Preference Alignment via In-context DPO. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 161–178, Albuquerque, New Mexico. Association for Computational Linguistics.